Tensorflow with AirPrime WP7702

Hi everyone,

I would like to include a third-party package allowing me to perform machine learning with an AirPrime module (For now I am focusing on the Tensorflow Framework with a WP7702 on a mangOH board).

The filesystem being an ubifs I can not install the python packages in a “classic way”. For that I have to modify the Yocto source code. (AFAIK).

By looking to the OpenEmbedded layer index I stumbled upon differents layers with recipes I could use (see for example meta-machinelearning).

However, those layers are only available with a version of Yocto >= 2.6 (thud).
The last release of the WP7702 firmware is on Yocto 2.5.3.

Does someone know if this module will have a firmware update with a more recent version of Yocto ? If yes, when ?
Does someone ever tried to include deep learning within an AirPrime module ?

Also, I saw that on gitlab, BSP (meta-swi) have a branch for Yocto >= 2.6.
When looking for conf files, i could find this:

#@TYPE: Machine
#@NAME: swi-mdm9x28-wp
#@DESCRIPTION: **Machine configuration for swi-mdm9x28 wp76xx and wp77xx systems**
require conf/machine/swi-mdm9x28.conf

# Beginning of the WP recipes

# Add a generic override for WP and the initial machine without WP
OVERRIDES_append = ":swi-mdm9x28-wp:swi-mdm9x28"

Can I replace this layer with the one downloaded from Sierra Source to have right software/tools on my distro?

Any help would be higly appreciated,
Thanks !